arX

iv:1

810.

0655

7v2

[ph

ysic

s.ac

c-ph

] 3

1 Ja

n 20

19

Accelerator Design for the CHESS-U Upgrade

J. Shanks,∗ J. Barley, S. Barrett, M. Billing, G. Codner, Y. Li, X. Liu, A.

Lyndaker, D. Rice, N. Rider, D.L. Rubin, A. Temnykh, and S.T. WangCLASSE, Cornell University, Ithaca, New York 14853, USA

(Dated: February 1, 2019)

During the summer and fall of 2018 the Cornell High Energy Synchrotron Source (CHESS) isundergoing an upgrade to increase high-energy flux for x-ray users. The upgrade requires replac-ing one-sixth of the Cornell Electron Storage Ring (CESR), inverting the polarity of half of theCHESS beam lines, and switching to single-beam on-axis operation. The new sextant is comprisedof six double-bend achromats (DBAs) with combined-function dipole-quadrupoles. Although theDBA design is widely utilized and well understood, the constraints for the CESR modificationsmake the CHESS-U lattice unique. This paper describes the design objectives, constraints, andimplementation for the CESR accelerator upgrade for CHESS-U.

I. INTRODUCTION

The Cornell Electron/positron Storage Ring (CESR)was designed as an electron-positron collider, with twodiametrically opposed interaction regions. The layout ofthe storage ring is constrained by the very nearly circulargeometry of the tunnel that it shares with the existingsynchrotron. The CLEO detector required a long straightsection to accommodate strong IR quads for a mini-betainsert and anti-solenoids to compensate the 1.5 T CLEOsolenoid, in addition to the detector itself. Hard-benddipoles were required to bend the trajectory into the longstraight, with gentler bends immediately adjacent to thestraight to mitigate synchrotron radiation background inCLEO. The collider layout and characteristic optics areshown in Fig. 1.

The Cornell High Energy Synchrotron Source(CHESS), founded contemporaneously with CESR, wasinitially comprised of three x-ray extraction lines off thehard-bend dipoles from the counter-clockwise electronbeam, expanding in 1988-89 with the addition of fourbeam lines (including one permanent-magnet wiggler)off the clockwise positron beam in CHESS East, andagain in 1999 with the construction of G-line, addingthree end stations fed by an insertion device from thepositron beam.

Following the conclusion of the CLEOc HEP programin 2008, CESR was reconfigured for the CESR Test Ac-celerator program (CesrTA) [1] damping ring R&D pro-gram. The CLEO particle detector drift chamber wasremoved and the final focus replaced with a conventionalFODO arrangement. The long straight was outfittedwith six superconducting damping wigglers, relocatedfrom two straights in the arcs.

The CesrTA layout also enabled the operation ofCHESS in the arc-pretzel configuration with counter-rotating beams of electrons and positrons [2]. Layoutand optics are shown in the middle plots in Fig. 1.

∗shanks@cornell.edu

CHESS was designed to run parasitically with HEPoperations, with beam lines fed by both electron andpositron beams. Running with two electrostatically-separated beams greatly complicates many facets of oper-ation, including restricting bunch patterns and maximumper-bunch current, increasing the tune plane footprint,and impairing the dynamic and momentum apertures.With the conclusion of HEP operations, CESR

is being reconfigured as a dedicated high-energy,high-performance single-beam synchrotron light source,CHESS-U. The conversion to operation with a singleclockwise beam requires rebuilding the beam lines fed bythe counter-clockwise beam. The new layout eliminatesthe long IR straight and hard bends in favor of evenlydistributed achromats and insertion straights, as shownin Fig. 1.

II. LATTICE DESIGN

A. Design Objectives and Constraints

The accelerator requirements for the CHESS-U up-grade are: 1) Minimize the natural emittance for a singleon-axis beam at 6.0 GeV; 2) Introduce straights for inde-pendent insertion devices for most end stations; 3) Pre-serve compatibility with existing injector system (linacand synchrotron booster ring); and 4) Increase high-energy x-ray flux for all end stations.Most of the CHESS beam lines are presently located

in the L0 main experimental hall, fed by both electronand positron beams, see Fig. 2. Eight end stations arelocated in L0, four from each beam; a further three endstations were added in 1999 as an excavation out of thewest CESR tunnel, fed by the clockwise-oriented positronbeam. For single-beam operation, a clockwise orientationwas chosen in order to preserve the existing six end sta-tions from the F-line and G-line facilities.The existing storage ring layout in L0 includes one long

straight section for the former CLEO detector, which isnot optimal for low-emittance x-ray light source oper-ation. The present optics do not have enough degrees

2

FIG. 1: At left is the magnetic layout of the L0 arc of the storage ring; top) Colliding beams, middle) CesrTA andCHESS arc-pretzel, bottom) CHESS-U. At right optics in the L3 straight for colliding beams, arc-pretzel and

CHESS-U.

of freedom to constrain the dispersion in the hard benddipoles which bring CESR into the long straight, and asa result, the hard bends dominate the natural emittance(see Fig. 3). Additionally, the long straight section can-not accommodate multiple source points for x-ray beamlines, which require angular separation between sources.

The L0 experimental hall was also home to the three-story CLEO particle detector, which ran for HEP oper-ation from 1979-2008. For single-beam CHESS opera-tion, several x-ray extraction lines would pass throughthe CLEO iron; to make room for new source points, theCLEO detector was disassembled and removed in 2016.

B. Linear Optics

In order to maximize the number of end stations, a lay-out was chosen with six 13.8 m double-bend achromatswith 3.5 m insertion device (ID) straights. It is worthnoting the cell length is approximately half that of manythird-generation storage ring light sources. DBAs werechosen over a multi-bend achromat design for a numberof reasons: 1) The space available is extremely limited; 2)The dynamic aperture requirements would be incompati-ble with the existing accumulation injection scheme; and3) the cost and complexity would be beyond the scope ofthis project.

The new DBAs utilize 2.35 m defocusing combined-function sector dipole-quadrupoles (DQs). Though thebend radius of the new CHESS-U dipoles is nearly iden-tical to the old hard bends (31.4 m vs. 31.7 m), thedispersion is well suppressed in the new achromats, lead-

ing to a factor of four reduction in the global emittancewhile only replacing one-sixth of the storage ring.

A single DBA shown in Fig. 4. The natural emittancefor a single CHESS-U cell is 2.56 nm·rad at 6.0 GeV.Matched into the remainder of the CESR, the full ringoptics are shown in Fig. 5. Linear and nonlinear opticswere optimized using Tao [3], which is built on the Bmadaccelerator library [4].

The six new achromats span the 100-meter region be-tween the two superconducting rf straights, and beginsand ends with straights for insertion devices. Five ofthe seven straights will be used for CHESS IDs; thereis not presently space for end stations for the remain-ing two straights. The remaining two straights will haveone CESRc superconducting damping wiggler each [5],to be used only in low-energy machine studies. The newCHESS-U lattice will enforce zero dispersion through therf cavities, a condition which is not met in present two-beam CHESS optics. The new achromats are matchedinto the existing FODO structure with four additionalquadrupoles in each of the rf straights.

The storage ring energy will be increased to 6.0 GeV.Aside from CLEOc HEP operation in 2003-2008 and oc-casional CesrTA machine studies from 2008-2016, whenthe energy was lowered to 2.085 GeV/beam, CESR hasoperated at 5.3 GeV/beam. CHESS operation has beenexclusively at 5.3 GeV. CESR was originally designedfor an energy range of up to 8.0 GeV, though the high-est operating energy to-date is 5.6 GeV. Machine stud-ies conducted in 2017 in the present CESR configurationdemonstrated injection and accumulation at 6.0 GeV.Every quadrupole and sextupole in CESR is indepen-

3

(a) Layout before CHESS-U.

(b) CHESS-U layout.

FIG. 2: CHESS layout, before and after CHESS-U upgrade.

dently powered, allowing significant flexibility in operat-ing conditions. Therefore the remaining 5/6 of the stor-age ring requires no modification.CHESS-U lattice parameters are summarized in Ta-

ble I.

C. Nonlinear Optics

Due to the lack of periodicity in CESR (the nearestsymmetry in the ring is a mirror symmetry about theNorth-South axis), nonlinear optics are numerically op-timized for CHESS-U using 76 independently-controlledsextupoles with an implementation of J. Bengtsson’s res-onance driving term (RDT) formalism [6] in Bmad. Asthere is no symmetry, all sextupoles are allowed to varyseparately.An unusual design feature of the CHESS-U achromat

is a lack of sextupoles in the new sextant. The peakdispersion in the DBAs is around 20 cm, whereas the av-erage dispersion in the remaining 5/6 of CESR is around1 m; the betas in the new achromats are also suppressedcompared to the rest of the machine. As such, any sex-tupoles in the new sextant have little leverage on thechromaticity. Optimizations showed the inclusion of har-monic sextupoles in the new DBA cells did not signifi-cantly improve the nonlinear optics, and were thereforeomitted.

The new DQ magnets have a significant b2 multipole,contributing roughly ∓0.5 to the horizontal and verticalchromaticity, respectively. This was compensated duringthe RDT reduction optimization.

4

FIG. 3: Element-by-element contribution to the naturalemittance, via the I5 radiation integral. The greenhatching indicates the region to be upgraded for

CHESS-U. s = 0 corresponds to the center of the formerCLEO interaction region, which is also the center of the

CHESS user region of the storage ring. The largestcontributions to I5 in the existing CESR layout

correspond to the hard bend dipoles which bend intothe long CLEO IR straight.

FIG. 4: One DBA cell for the CHESS-U upgrade. Blueboxes are DQs, purple are quadrupoles, green aresteerings. Red ticks are beam position monitors.

III. LATTICE CHARACTERIZATION

A. Performance

Four of the five sectors with CHESS IDs will havecanted 1.5 m CHESS Compact Undulators (CCUs) [7–9].Sector 1 will use the existing 24-pole F-line wiggler [10].Anticipated pinhole flux was simulated with SPECTRA[11], shown in Fig. 6.

B. Error Tolerance

Error tolerance for CHESS-U was assessed using thetools and low-emittance tuning methodology developedfor the CesrTA program [13]. The optics correction pro-

FIG. 5: Full ring optics for CHESS-U. s = 0corresponds to the center of the former CLEO

interaction region, which becomes the center of theSector 4 insertion device straight. The green hatchingindicates the region being upgraded for CHESS-U.

Parameter CHESS CHESS-U UnitsCircumference 768.438 768.438 m

Energy 5.289 6.0 GeVSpecies e+ and e− e+ –Current 120/120 200 mA

ǫx 98 29.6 nm·radEmittance coupling 1% 1% –

βx,y at IDs 7.9, 3.1 11.2, 2.6 mηx at IDs 0.42 0 m

IDs 3 9 –End Stations 11 12 –

Qx,y 11.28, 8.78 16.55, 12.63 –Q ′

x,y -15.95, -14.20 -25.64, -26.76 –αp 9.2× 10−3 5.7× 10−3 –

Bunch Length 16 14 mmPer-bunch Current 7 2.2 mATouschek Lifetime >24 40 hrs

RF Voltage 5.2 6 MV

TABLE I: Lattice parameters for CHESS today andCHESS-U upgrade.

FIG. 6: SPECTRA modeling of flux through a 1 mm2

pinhole at 20 meters, before and after CHESS-Uupgrade [12]. “F-line wiggler” refers to the 24-pole

wiggler for Sector 1.

5

100 101 102 103 104εy [pm]

010203040

No. S

eeds

iter 0iter 1iter 2iter 3

10−3 10−2 10−1ηy [m]

010203040

No. S

eeds

iter 0iter 1iter 2iter 3

10−2 10−1Cbar12

010203040

No. S

eeds

iter 0iter 1iter 2iter 3

FIG. 7: Statistical analysis of 100 misaligned andcorrected lattices. Iter0 corresponds to the misalignedlattice before correction; the following three iterationscorrespond to the steps listed above. 95th-percentile

ǫy = 4.1 pm after the final correction.

cedure used here is identical to that in CesrTA:

1. Measure closed orbit; correct using horizontal andvertical steering magnets.

2. Measure betatron phase advance, local coupling[14], and horizontal dispersion; correct usingquadrupoles and skew quadrupoles.

3. Measure orbit, local coupling, and vertical dis-persion; correct using vertical steerings and skewquadrupoles.

Though the new DQs have trim windings, the correc-tion simulation did not allow for the DQs to vary. Erroramplitudes are listed in Tables V and VI, and representthe best estimate of anticipated alignment and BPM er-rors.A summary of the misalignment study is shown in Fig.

7. The statistical impact of misalignments and correc-tions on the Twiss parameters is shown in Fig. 8. The95th-percentile vertical emittance due to optics errors af-ter correction is 4.1 pm.

C. Dynamic and Momentum Aperture

Dynamic and momentum aperture were evaluated viafrequency map analysis [15]. Particles were tracked fora total of 2048 turns at each initial amplitude; dynamicaperture plots were normalized to beam sigmas, assuming1% emittance coupling.

FIG. 8: Statistics of machine optics distortion as afunction of longitudinal position for 95 out of 100misaligned and corrected lattices (representing95th-percentile anticipated results). Black curveindicates the average over 100 seeds, blue shading

indicates the standard deviation, and the red dashedlines indicate the maximum and minimum excursion atany given point in the lattice. The green hatched regionindicates the region with new CHESS-U achromats,

where betas and dispersion are suppressed compared tothe rest of the ring.

Systematic multipoles for dipoles, quadrupoles, andsextupoles were determined via field modeling or mea-surement, and are listed in Table VII. Due to the com-pact nature of the CCU magnet arrays, the field roll-offis significant: vertical field is reduced 10% at a 10 mmhorizontal displacement. Systematic multipoles and thefield roll-off and measured field integrals for CCUs wereincluded in some tracking simulations [2], denoted in therelevant figure captions.Dynamic aperture and its footprint in the tune plane

are shown in Fig. 9 for combinations of systematic mul-tipoles, insertion devices, and misalignment/correction;corresponding momentum aperture with systematic mul-tipoles, IDs, and misalignment/correction are shown inFig. 10. The color scale in these plots is the tune diffu-sion, given by

dQ = log√

∆Q2x +∆Q2

y (1)

D. Injection Simulation

The injection scheme for CESR relies on accumula-tion from a 60 Hz synchrotron booster ring. Injectionefficiency was simulated via multi-particle tracking. 200macroparticles were launched with a phase space match-ing the beam at the end of the synchrotron booster ex-traction line. Particles which survive for 400 turns wereconsidered captured. Details are found in [2].Results of the injection tracking simulations are shown

in Fig. 11. It should be noted that the injection efficiency

6

−30 −20 −10 0 10 20 30x/σx

0

5

10

15

20

25

30

y/σ y

−12

−10

−8

−6

−4

−2

0.535 0.540 0.545 0.550 0.555Qx

0.624

0.626

0.628

0.630

0.632

0.634

0.636

Qy

(a) Ideal lattice.

−30 −20 −10 0 10 20 30x/σx

0

5

10

15

20

25

30

y/σ y

−12

−10

−8

−6

−4

−2

0.535 0.540 0.545 0.550 0.555Qx

0.624

0.626

0.628

0.630

0.632

0.634

0.636

Qy

(b) With systematic multipoles and realistic undulator field tracking.

−30 −20 −10 0 10 20 30x/σx

0

5

10

15

20

25

30

y/σ y

−12

−10

−8

−6

−4

−2

0.535 0.540 0.545 0.550 0.555Qx

0.624

0.626

0.628

0.630

0.632

0.634

0.636

Qy

(c) With systematic multipoles, realistic undulator field tracking, and misalignments and corrections (95th-percentile seed).

FIG. 9: Dynamic aperture via the frequency map (left column), and its associated footprint in the tune plane (rightcolumn). Color scale indicates tune diffusion, defined in Eqn. 1. Red dashed line indicates the minimum projected

aperture.

is not expected to reach 100% as there are two 10 mmfull-aperture vertical collimators used to mask againstparticle losses on the small-gap permanent-magnet IDs.

IV. MAGNET DESIGN

Most constraints on magnet design are more or lessstandard. The specific design impacts on the CHESS-Umagnets are discussed here. Details are available in [16].

A. Combined Function Dipole-Quadrupoles

Space constraints and emittance reduction motivatedimplementing DQs in the new DBA cells rather than con-ventional dipoles. The DQs have a magnetic length of2.35 m, bend radius of 31.4 m, and gradient -8.77 T/m.The full-gap aperture along the design trajectory is38.4 mm. The 3D OPERA model and excitation curvefor the DQ magnet are shown in Fig. 12. Although themagnet yoke is rectangular in geometry, the pole piecesfollow a sector dipole trajectory.

7

−30 −20 −10 0 10 20 30x/σx

−10

−5

0

5

10

δ/σ E

−12

−10

−8

−6

−4

−2

0.5400 0.5425 0.5450 0.5475 0.5500 0.5525 0.5550 0.5575Qx

0.622

0.624

0.626

0.628

0.630

0.632

0.634

Qy

FIG. 10: Momentum aperture via the frequency map and its associated footprint in the tune plane, with systematicmultipoles, realistic undulator field tracking, and misalignments and corrections (95th-percentile seed). Color scale

indicates tune diffusion, defined in Eqn. 1.

0.550 0.575 0.600 0.625 0.650 0.675Qx

0.58

0.60

0.62

0.64

0.66

0.68

0.70

Qy

0

20

40

60

80

100

FIG. 11: Injection tune scan simulation with systematicmultipoles, realistic undulator field tracking, and

misalignments and corrections (95th-percentile seed).Color scale indicates capture efficiency in percent.

B. Quadrupoles

All achromat quadrupoles are horizontally-focusing, asthe vertical focusing is done by the DQs. There are twofamilies of quadrupoles, with magnetic lengths 0.400 mand 0.362 m, both with the same pole profile; the pri-mary distinctions are length and the extension of theiron to allow for x-ray extraction on one of the two fam-ilies. Both quadrupole families have maximum gradient39 T/m, with a bore radius of 23 mm. Fig. 13 showsthe magnetic field in one quarter of a non-extractionquadrupole and the excitation curve; an extraction quadmodel is shown in Fig. 14.

FIG. 12: Top: Magnetic field in one half of a DQ iron.Peak field in the iron is approximately 2.3 T. Bottom:

DQ excitation curve.

C. CHESS Compact Undulators

Four of the five new sectors will be equipped withcanted pairs of CHESS Compact Undulators (CCUs).The design for the CCUs is detailed elsewhere [7–9]. Anend-view of the CCU profile is shown in Fig. 15a; per-spective CAD drawing of CCU assembly is shown in Fig.15b. Parameters for the CCUs are summarized in Table

8

FIG. 13: Top: Magnetic field in one quarter of anon-extraction quadrupole iron. Peak field in the iron isapproximately 2.3 T. Bottom: Quadrupole excitation

curve.

FIG. 14: Extraction quadrupole model.

II.The CCUs utilize a variable-phase design rather than

variable-gap in order to minimize the supporting infras-tructure. As implemented in CESR the CCUs have a7.0 mm out-of-vacuum vertical aperture between poles,mandating the vacuum chamber to fit inside this aper-ture. The associated vacuum chamber for the CCUs isdescribed in Section VB.

(a) End-on view.

(b) Perspective view.

FIG. 15: CAD model of the Cornell CompactUndulator (CCU).

Parameter ValueLength 1.477 mPeriod 28.4 mm

Peak field 0.952 TPole gap 7.0 mm

First harmonic 2.9 keV

TABLE II: Parameters for the CHESS CompactUndulators.

CHESS has operated with two 1.5 m CCUs in a cantedstraight since 2014, with two counter-rotating beams.Total beam current has routinely exceeded the CHESS-U target 200 mA in single-beam positron tests at thepresent CHESS beam energy of 5.3 GeV.

9

D. Power Distribution

For the new CESR south arc, the magnet power dis-tribution was chosen to be be compatible with existinginfrastructure. There are two main pieces to this: theCESR dipole circuit and the CESR quadrupole bus sys-tem. The controls for both are derived from a uniquecomputer bus known as the Xbus [17, 18]. The Xbushandles all digital and analog controls, read back signals,and interlocks associated with the operation of the mag-net power supplies. All magnets can be controlled indi-vidually or collectively as needed from the central controlroom.The CESR Dipole circuit is a series circuit of approx-

imately 120 bend magnets driven by two 0.5 MW Tran-srex power supplies at 500 V and 740 A to obtain 6 GeV.The series chain is made up of normal bend, soft bend,and hard bend dipole magnets. The new DQ magnetswill be part of this existing circuit.The operation of CHESS-U requires five quadrupole

buses: East, South East, South Center, South West, andWest. The SE, SC, and SW are new buses installed ex-pressly for the south arc installation. The West and Eastquad buses are preexisting. Each quad bus is driven byan 80 V 1125 A EMHP power supply. These buses sup-ply a fixed 70 V primary voltage to a large number ofswitching power supplies (DC-to-DC converters). Thesepower supplies excite all quadrupole, sextupole, steering,skew quadrupole, skew sextupole, and octupole magnets.Through the Xbus, each switching power supply can becontrolled independently, giving flexibility for tuning op-tics. The current draw depends on the optics, typicallyscaling with beam energy.

E. Support and Alignment

Each new CHESS-U cell was split into three sections:an upstream girder, a downstream girder, and an inser-tion device straight. The two girders support all achro-mat magnets, each holding two quadrupoles and one DQ,plus corrector magnets. The insertion device straight issuspended from above in order to allow access to the up-stream sector’s front end. One sector is illustrated in Fig.16.Magnetic alignment of the DQ and quadrupoles is

achieved using a combination of vibrating wire [19] andHall probe. The procedure is as follows:

1. Establish vibrating wire (VW) position with re-spect to girder fiducials

2. Establish Hall probe position with respect to VW

3. Place wire on beam axis

4. Align Q1 magnetic axis with VW

5. Align Q2 magnetic axis with VW

FIG. 16: Rendering of one sector for CHESS-U from theradial-outside. Beam travels clockwise (right-to-left).

FIG. 17: Setup for magnetic centering of DQ andquadrupoles on a girder. [16]

6. Align DQ using Hall probe

The setup used for this alignment procedure is dia-grammed in Fig. 17. Using this method, the magneticcenters of quadrupoles and the DQ on one girder arealigned to within 32 µm.

V. VACUUM DESIGN

A. Design Considerations

As a part of particle beam transport system, theCHESS-U vacuum chambers need to provide adequatephysical beam aperture, while maintaining sufficientclearance to all CHESS-U magnets. The vacuum pump-ing system is designed to achieve and to maintain ultra-high vacuum (UHV) conditions, with an average vacuumpressure< 1 nTorr with stored 200 mA beam at 6.0 GeV.Additionally, all vacuum chambers must be designed tomanage heating from synchrotron radiation (SR) gener-ated by the bending magnets, with a minimum factor ofsafety (FOS) >2. The vacuum chambers will also housea suite of BPMs for beam instrumentation.

B. Vacuum Chambers

All CHESS-U vacuum chambers are constructed fromtype 6061 and 6063 aluminum alloys. Similar to themodular magnet design concept, the CHESS-U vacuumchambers conform to the periodic achromat cells. Owing

10

FIG. 18: A typical CHESS-U Achromat Cell vacuum chamber string is consisted of three long vacuum chambers.

to the limited longitudinal spaces between the magnets,the vacuum string for each achromat is made into threeflanged chambers, as shown in Fig. 18. More specifically,each achromat is comprised of two long dipole chambers,a 4.91 m long downstream “Dipole C” chamber and a5.03 m long upstream “Dipole A” chamber, which areintegrated with magnet girders, and a 3.69 m long undu-lator vacuum chamber.

The majority of the chambers are made from 3 stylesof 6063 aluminum extrusions (Fig. 19): dipole extrusion,quad extrusion and undulator extrusion. The minimumclearances between magnet pole tips and the extrusionsare slightly less than 2 mm, as illustrated in Fig. 20.The dipole and quad extrusions have a beam apertureof 52 mm (H) × 22 mm (V), while the vertical aper-ture in the undulator extrusion is 5 mm. The dipole andundulator extrusions have ante-chambers for distributedpumping.

The bending section of a dipole C chamber, as shownin Fig. 21a, is made from the dipole-style extrusion (Fig.19b). Precision bending (31.17 m radius, 4.26 degreebending angle) of the extrusion was achieved via thestretch-forming technique, in which an aluminum extru-sion was formed to a precision die while being stretchedto its yield stress. The stretch-formed section was thenmachined to its design form, and the quad-style (Fig.19a) extrusions welded to both ends. Four sets of BPMswere welded directly onto the quad extrusions near loca-tions of quadrupole magnets. Other functional vacuumcomponents, including an RF-shielded sliding joint andpumping ports, were also welded to the quad extrusions.The chamber completes with a pair of stainless flangeswith aluminum-to-stainless steel bi-metal transitions.

The dipole A chamber design is much more compli-cated than the dipole C chamber, as it needs to provideegress for the X-ray from the canted undulators. Thedipole A design is illustrated in Fig. 21b. To incorporatethe compact geometry needed for the passage of bothstored electron beam and egress of X-ray from the undu-lators, the chamber main body is constructed by weldingof two precision machined “clam-shells” of 6061-T6 alu-minum alloy. Cooling channels and NEG strip mountingfeatures are also machined in the main body. Similarto the dipole C chamber, quad extrusions are welded toboth ends of the main body. Two sets of BPMs, oneRF-shielded sliding joint and multiple pump ports arealso directly welded to the chamber. At separation junc-

FIG. 19: Three types of aluminum extrusions used inconstruction of CHESS-U vacuum chambers. (A) Quadextrusions are used in straight sections with tight fitbetween magnet poles and beampipes; (B) Dipoleextrusions are used to form bending sections in DQ

magnets, which includes ante-chamber for NEG-strips.(C) Undulator extrusions are the base material for the

undulator vacuum chambers.

tion of the electron beam and the X-ray, an insertablecrotch absorber is mounted onto the dipole A chamber,to intercept up to 5.4 kW of SR power from the dipolemagnet, while allowing required passages of both elec-tron beam and the X-ray. The crotch absorber is oneof the most challenging CHESS-U vacuum components,due to very limited available space. Its design is based onthe existing CHESS crotches, consisting of a pure beryl-lium ring (5 mm thick) and a water-cooled copper core.The beryllium ring, vacuum brazed to the copper core,dilutes the SR power density by bulk absorption and scat-tering along its depth. FEA analysis indicates that thecrotch absorber has sufficient margin of safety to handlethe SR power from the design CHESS-U beam parame-ters (6.0 GeV, 250 mA). Design and construction of theCHESS-U undulator chambers are similar to an existingCCU chamber in operation for the last four years. As

11

(a) Dipole extrusion in a DQ magnet.

(b) Quadrupole extrusion in a quadrupole magnet.

FIG. 20: Cross-sections of extrusions in their respectivemagnets.

shown in Fig. 21c, the undulator beampipe is made fromthe CHESS-U undulator extrusion (Fig. 19c). Two sec-tions (1.51 m each) of the extrusion are machined to athickness of 0.61 mm to accommodate the 7.0 mm CCUpole gap. The machining of the undulator extrusion isdone dry to avoid difficulty in cleaning the long beampipewith 5 mm vertical aperture with thin walls. The wallthickness at the thin section was constantly checked usingultrasonic thickness gauges during the machining. Twoend assemblies are welded to the undulator beampipe,that gently transition vertically from 5 mm to 22 mm,the nominal CHESS-U vertical aperture. Six pumpingports are welded to the ante-chamber of the undulatorextrusion.

C. Vacuum Pumping and Performance Simulation

All CHESS-U vacuum chambers were baked to 150◦ Cand then back-filled with ultra-high purity nitrogenthrough a MATHESON NANOCHEM R© purifier thatreduces H2O/HC level below ppb level. A post-installation 95◦ C hot water bakeout of all CHESS-Uchambers is also planned. With these measures, SR-induced desorption (SRID) is expected to be the dom-inant source of gas load.Owing to distributed nature of the SRID gas load and

restricted conductance of the CHESS-U beampipe, it isessential to have distributed vacuum pumping. Non-evaporable getter (NEG) strips (st-707 SAES Getters)

(a) Dipole C vacuum chamber.

(b) Dipole A vacuum chamber.

(c) Undulator vacuum chamber.

FIG. 21: Vacuum chamber detail.

were chosen for the distributed pumps in the dipolechambers over distributed ion pump in CESR, due tothe limited space available. The structure of the NEGstrip assembly is shown in Fig. 22. A 30 mm wide NEGstrip is supported on a stainless steel ribbon with peri-odic stainless steel clips. The clips are pinched to theNEG strip, and are riveted to the stainless steel ribbonthrough sets of alumina spacers. The NEG strip will beactivated by resistive heating to 500◦ C with a DC powersupply. Flexible connections at both ends of the NEGstrip to the electric vacuum feedthroughs allow thermalexpansion during activation.

12

FIG. 22: - NEG strip assembly in dipole chambers. (A)30-mm NEG (SAES Getters st-707) strip; (B) Stainless

steel carrier ribbon; (C) Stainless steel clips; (D)Alumina spacers.

Very high SRID gas load is expected in the initial beamconditioning stage. Simulations showed that the NEGstrips may not have sufficient pumping capacity duringthe CHESS-U commissioning phase, and it can be dis-ruptive to the initial operation with frequent activationcycles. Therefore, compact high capacity NEG modularpumps, CapaciTorr Z200 and NexTorr Z100 (SAES Get-ters), are installed to aid the initial beam conditioning.To handle potential noble (or non-gettable) gases, a largesputtering ion pump (approx. 110 l/s) is also installedon each achromat cell.The CHESS-U vacuum pumping performance was

evaluated using MolFlow+, a Test-Particle Monte-Carlosimulator developed at CERN. The simulations showedthat the installed vacuum pumping system is capableof achieving required level of vacuum for CHESS-U op-erations. Figure 23 displays an example of simulatedpressure profile in one CHESS-U achromat cell, after100 Amp-hr beam conditioning, achieving an averagepressure of ≈ 1.8 × 10−9 torr. Pressure bumps are dueto lack of pumping in the straight sections populated bythe magnets.

VI. INSTRUMENTATION

CESR is well-instrumented for a multitude of beam-based measurement techniques, thanks to the CesrTAprogram. Details are documented elsewhere [20]. A fewhighlights are noted here.

A. CESR Beam Position Monitors

CESR is presently equipped with 110 peak-detectionBPMs of an in-house design (CBPM-II), capable of mea-suring bunch-by-bunch turn-by-turn positions for bunchspacings ≥ 4 ns. 100 CBPMs are used for routine op-tics correction and orbit monitoring. Each achromat in

FIG. 23: Simulated CHESS-U pressure profiles with250 mA stored beam with 100 A-hr of

beam-conditioning, assuming NEG pumps at nominalpumping speed, using MolFlow+, which showed an

average pressure of ≈ 1.8× 10−9 torr.

Parameter SpecificationFront-end bandwidth 500 MHz

Absolute position accuracy (long-term) 100 µmSingle-shot position resolution 10 µmDifferential position accuracy 10 µm

Channel-to-channel sampling time resolution 10 psec

TABLE III: CBPM-II specifications.

CHESS-U will be instrumented with four CBPM-II mod-ules, one adjacent to every quadrupole. CBPM-II specifi-cations are listed in Table III; further details are availablein [21].

With the installation of canted CCU undulators in2014 [22], three Libera Brilliance Plus processors havebeen used for continuous turn-averaged position moni-toring at the existing 3.5 m-long 4.6 mm-aperture CCUchamber.

B. X-Ray Video Beam Position Monitors (vBPMs)

Each x-ray front end will be equipped with twovBPMs, one for each extracted beam, to provide hor-izontal and vertical position. The vBPMs will use anedge-on diamond blade monitor which was prototyped onA-line in 2015 [23], rendered in Fig. 24. The diamond-blade style vBPM allows for a compact design close tothe source (roughly 10 m), where the canted undulatorbeams are only separated by 10 mm.

The vBPMs are presently used in conjunction withLibera BPMs around the small-gap undulator chamber tomonitor x-ray trajectories in real-time and correct posi-tions after every topoff, typically every five minutes. Thisstrategy will be expanded in CHESS-U, where each sectorwill have two Libera Brilliance Plus monitors dedicatedto real-time position monitoring for position feedback.

13

(a) Schematic of diamond bladesintercepting synchrotron radiation. Theblades are separated by approximately

7 mm.

(b) Implementation of two adjacent vBPMsfor monitoring positions of two cantedundulator beams separated by 10 mm

(1 mrad angular separation).

FIG. 24: vBPM developed for A-line.

C. Turn-by-Turn Feedback

Bunch-by-bunch turn-by-turn horizontal, vertical, andlongitudinal feedback is accomplished using three Dim-tel feedback front-end and processing modules [24]. Thelow level outputs are connected to 200 W, 150 MHz ENIamplifiers which drive horizontal and vertical striplinekickers. The longitudinal processor drives a 200 W solid-state RF amplifier that in turn drives a 1 GHz low Qcavity kicker. While the system can handle bunch spac-ings as short as 2 ns, gain falls off because of the shortinterval required to provide feedback. A bunch spacing of14 ns provides for the maximum kick as well as matchingbetter with the injector RF frequency.

VII. BUNCH PATTERNS

Historically, bunch patterns were constrained by thebeam-beam interaction from electrostatically-separatedcounter-rotating beams in the same vacuum chamber.With the simplification of single-beam operation, there

are many more possibilities.There is increasing interest in time-resolved experi-

ments (see, for example, [25, 26]), and a number of pos-sible bunch patterns are under consideration. The con-straints on bunch patterns are now discussed, and severalbunch patterns proposed.

A. Mode Coupling Instability Threshold

Many light source upgrades are favoring mult-bendachromat designs with strong focusing and small vacuumapertures (see, for example, [27, 28]). However, the re-sistive wall impedance scales as 1/r3, quickly limitingthe single-bunch current due to mode coupling instabil-ity [29]. The vacuum chamber aperture for CHESS-Uhas been kept relatively large (52 mm × 22.5 mm full-aperture) in an effort to keep the maximum single bunchcurrent as high as possible.The transverse mode coupling instability (TMCI)

threshold has been estimated for CHESS-U based on an-alytic resistive wall calculations [30] and detailed T3Pmodeling for discontinuities [31, 32]. The limit is esti-mated to be around 25 mA (64 nC) [33]; conservatively,only bunch patterns up to 20 mA are considered.The present single-bunch current limit is 10 mA

(26 nC) per bunch, to protect instrumentation (BPMmodules and feedback amplifiers). Options for raisingthe limit are under discussion, opening the path to highbunch-current operating modes. These options includepreventing damage to the feedback amplifiers by limitingbeam-induced power transmitted back to the amplifier.

B. Momentum Aperture and Touschek Lifetime

The momentum compaction factor αp for CHESS-Uis rather large compared to other modern light sources,resulting in a limited momentum aperture (Fig. 25, top).The Touschek lifetime (Fig. 25, bottom) at the designemittance coupling of 1%, total RF voltage of 6 MV, anddesign bunch current of 2.22 mA (in 18 × 5 mode) isroughly 40 hrs.Extrapolating to 25 mA, the anticipated Touschek life-

time at 0.1% coupling is approximately one hour; at 1%coupling, this increases to around 3 hours.

C. Electron Cloud

Initial commissioning will be done with positrons. Assuch, bunch patterns will be limited by tune shift andhead-tail instability due to the electron cloud effect [34].Each of the bunch patterns proposed here were exam-ined for electron cloud build-up and trapping [35]. Theelectron cloud tune shift was constrained to be less thanor equal to the electron cloud tune shift in present two-beam bunch patterns (around 3.5 kHz, or ∆Q ≈ 0.009 in

14

(a) Momentum aperture.

(b) Touschek lifetime.

FIG. 25: Momentum aperture and Touschek lifetime asa function of RF voltage, bunch current, and emittance

coupling.

Parameter 18× 5 9× 5 9× 1# Trains 18 9 9

Train Spacing [ns] 84 224 280# Bunches / Train 5 5 1Bunch Spacing [ns] 14 14 n/aCharge/Bunch [mA] 2.22 4.44 20Total current [mA] 200 200 180

τTouschek at 1% Coupling [hrs] 40 17 4

TABLE IV: Bunch patterns for CHESS-U. 1 mA =1.6× 1010 particles = 2.56 nC.

both horizontal and vertical). It is possible that largertune shifts from electron cloud would be tolerable.

D. Proposed Bunch Patterns

Bunch patterns proposed for CHESS-U are listed inTable IV. Commissioning and early operation will be inthe 18 × 5 pattern, with the option of exploring otherbunch patterns as user interest dictates.In any of the proposed bunch patterns, CESR would

run with topoff. In present CHESS two-beam conditions,the topoff interval for positrons is 5 min, with a Touscheklifetime of approximately 12 hrs.

CESR will preserve the ability to store electrons in thecounter-clockwise direction for machine studies. Totalcurrent in the counter-clockwise direction will be limitedto 5 mA (13 nC) at 6 GeV, limited by synchrotron radi-ation heating on the sliding joints in the new achromats.This current limit will scale as E−4; therefore, the re-striction in the counter-clockwise direction at 2.1 GeV(frequently used for machine studies) will exceed the ex-isting administrative limit of 200 mA.

VIII. INSTALLATION AND COMMISSIONING

Installation of the CHESS-U upgrade began in June2018, and is scheduled to be completed in November2018. Beam commissioning will follow. All insertion de-vices are out-of-vacuum, which allows for initial commis-sioning and vacuum processing without IDs. Undulatorswill be installed one canted pair at a time, with a briefrecovery period after each installation to establish op-erating conditions. Beam will be delivered to all endstations by April 2019.

The positron beam presently circulates in the clockwisedirection; as such, initial commissioning and operationof CHESS-U will be with positrons, with the option toinvert the polarity of the injector and storage ring to runelectrons in the clockwise direction at a future date.

IX. SUMMARY

CHESS-U will deliver high flux photon beams at highenergy, enabling a diverse set of new experiments [12].The upgrade will improve the performance of the CESRstorage ring for x-ray production by almost an order ofmagnitude with CCUs and a factor of three for the 24-pole “F-line” wiggler. In turn, the transition to single-beam will drastically simplify operations. Bunch pat-terns compatible with timing mode are under considera-tion, pending evaluation during commissioning and ma-chine studies.

X. ACKNOWLEDGEMENTS

The authors wish to thank everyone involved in theCHESS-U upgrade project, and those who have con-tributed material to this paper.

Funding for the CHESS-U upgrade provided by NewYork State Capital Grant #AA737 / CFA #53676.

15

TABLE V: RMS error amplitudes for misalignment andcorrection simulation.

Error Dipole DQ Quad Sext Girder

x-offset [µm] 2000 100 200 300 100y-offset [µm] 900 100 110 300 100s-offset [mm] 2.3 2.0 5.2 5.2 0.1Pitch [µrad] 600 87 1100 1200 100Yaw [µrad] 300 87 62 800 100Roll [µrad] 500 500 500 500 100Gradient [%] – 0.1 0.1 0.1 –

TABLE VI: RMS BPM errors for misalignment andcorrection simulation.

Error AmplitudeTransverse offset [µm] 170Reproducibility [µm] 10

Roll [mrad] 12Timing [ps] 10Shear [µm] 100

Channel Gain [%] 0.5

Appendix A: Misalignments for Error Tolerance

Standard deviations for misalignments applied in errortolerance simulations are listed in Table V. The errorsfor dipoles, quadrupoles, and sextupoles are based onsurvey information. DQ and girder errors are determinedby extrapolation from survey information and magneticalignment characterization.BPM misalignments and errors are listed in Table VI.

Appendix B: Systematic Multipoles

The multipole expansion is parameterized as a frac-tional field error dB/B with normal (bn) and skew (an)components, where n = 1 corresponds to a quadrupole-like field. The reference radius for all multipoles listedin Table VII is 20 mm. All multipoles are normalized tothe main field.

TABLE VII: Systematic multipoles used in trackingsimulations. “CESR” magnets are those which areoutside the new CHESS-U achromats. Evaluated as

dB/B at a reference radius of 20 mm.

Element Class Multipole Amplitude

CESR Dipole b1 −0.5× 10−4

b2 −0.2× 10−4

b12 −2.4× 10−9

CESR Quadrupole b5 1.5× 10−4

b9 −0.053 × 10−4

CESR Sextupole b8 1× 10−4

CHESS-U Quads A/D b2 1± 6.5× 10−4

b3 0.4± 1.7× 10−4

b4 −0.3± 1.5× 10−4

b5 −4± 1.5× 10−4

a2 0.8± 7.8× 10−4

a3 −0.3± 2.8× 10−4

a4 −0.5± 1.9× 10−4

a5 0.1± 0.6× 10−4

CHESS-U Quads B/C b2 1.2± 6.1× 10−4

b3 0.8± 2.3× 10−4

b4 0.8± 1.8× 10−4

b5 −2.6± 2.1× 10−4

a2 −1.7± 3.3× 10−4

a3 −2.6± 4.0× 10−4

a4 −2.2± 2.9× 10−4

a5 0.3± 0.7× 10−4

CHESS-U DQ b2 −3.4± 5.6× 10−4

b3 −2.3± 1.9× 10−4

a1 −1± 2.8× 10−4

a2 −0.6± 2.6× 10−4

a3 0.1± 1.4× 10−4

16

[1] M. Billing, “The conversion of CESR to operate as theTest Accelerator, CesrTA. Part 1: overview,” J. Instrum.10, P07012 (2015), 10.1088/1748-0221/10/07/P07012.

[2] S. Wang, D. L. Rubin, and J.P. Shanks, “CESR Latticefor Two Beam Operations with Narrow Gap Undulatorsat CHESS,” in Proc. of International Particle AcceleratorConference (IPAC’16), Busan, Korea, May 8-13, 2016(2016) pp. 2968–2971.

[3] D. Sagan and J. C. Smith, “The TAO accelerator simu-lation program,” in Proceedings of the 2005 Particle Ac-celerator Conference, Knoxville, TN , edited by C. Horak(IEEE, 2005) pp. 4159–4161.

[4] D. Sagan, “Bmad: A relativistic charged particle simula-tion library,” Nucl. Instrum. Methods Phys. Res. A558,356–359 (2006).

[5] A. Mikhailichenko, “Optimized Wiggler magnet forCESR,” in 19th Particle Accelerator Conference (PAC2001) (2001) pp. 3648–3650.

[6] J. Bengtsson, The Sextupole Scheme for the Swiss LightSource (SLS): An Analytic Approach, Tech. Rep. SLS-TME-TA-1997-0009 (Swiss Light Source, Paul ScherrerInstitut, Villigen, Switzerland, 1997).

[7] A. Temnykh, D. Dale, E. Fontes, Y. Li, A. Lyndaker,P. Revesz, D. Rice, and A. Woll, “Compact undulatorfor the Cornell High Energy Synchrotron Source: Designand beam test results,” in Proceedings of the 11th Inter-national Conference on Synchrotron Radiation.

[8] A. Temnykh, A. Lyndaker, M. Kokole, T. Milhar-cic, J. Pockar, and R. Geometrante, “Construction ofCHESS compact undulator magnets at Kyma,” in Pro-ceedings of Advances in X-ray Free-Electron Lasers In-strumentation III .

[9] A. Temnykh, T. Kobela, A. Lyndaker, J. Savino,E. Suttner, and Y. Li, “Compact undulator forCornell High Energy Synchrotron Source,” IEEETrans. App. Superconductivity 22, 4100504 (2012),10.1109/TASC.2011.2181932.

[10] K. Finkelstein, “The new CHESS wiggler,” Rev. Sci. In-strum., 63, 305–308 (1992).

[11] T. Tanaka and H. Kitamura, “Spectra: A synchrotronradiation calculation code,” Journal of synchrotron radi-ation, 8, 1221–8 (2001).

[12] New Science Made Possible by CHESS-U , Tech. Rep.(CLASSE, Cornell University, 2016).

[13] J. Shanks, D. L. Rubin, and D. Sagan, “Low-emittancetuning at the Cornell Electron Storage Ring Test Accel-erator,” Phys. Rev. ST Accel. Beams 17, 044003 (2014),10.1103/PhysRevSTAB.17.044003.

[14] D. Sagan, “Betatron phase and coupling correctionat the Cornell Electron/positron Storage Ring,” Phys.Rev. ST Accel. Beams 3, 102801 (2000), 10.1103/Phys-RevSTAB.3.102801.

[15] J. Laskar, “Frequency analysis for multi-dimensional sys-tems. global dynamics and diffusion,” Physica D: Nonlin-ear Phenomena 67, 257 – 281 (1993).

[16] A. Temnykh, CHESS-U External Review, Tech. Rep.(CLASSE, Cornell University, Ithaca, NY, 2016).

[17] R. Littauer, S. E. Ball, R. G. Helmke, S. B. Peck, andD. H. Rice, “The cesr control system,” IEEE Transac-tions on Nuclear Science 26, 3275–3277 (1979).

[18] M. J. Forster, S. Ball, L. Bartnik, D. Bougie, R. Helmke,M. Palmer, S. Peck, D. Riley, R. Sholtys, andC. Strohman, “CESR control system upgrade to linuxhigh availability cluster,” in Proceedings of the 2012 In-ternational Particle Accelerator Conference, New Or-leans, LA (IEEE, 2012) pp. 3999–4001.

[19] A. Temnykh, “Vibrating wire field-measuring technique,”Nucl. Instrum. Methods A 399, 185–194 (1997).

[20] M.G. Billing, J.A. Dobbins, M.J. Forster, D.L. Kreinick,R.E. Meller, M.C. Rendina, N.T. Rider, D.C. Sagan,J. Shanks, J.P. Sikora, M.G. Stedinger, C.R. Strohman,H.A. Williams, M.A. Palmer, R.L. Holtzapple, andJ. Flanagan, “Instrumentation for the study of low emit-tance tuning and beam dynamics at CESR,” J. Instrum.12, T11006 (2017), 10.1088/1748-0221/12/11/T11006.

[21] M.G. Billing, W.F. Bergan, M.J. Forster, R.E. Meller,M.C. Rendina, N.T. Rider, D.C. Sagan, J. Shanks,J.P. Sikora, M.G. Stedinger, C.R. Strohman, M.A.Palmer, and R.L. Holtzapple, “Beam position monitor-ing system at CESR,” J. Instrum. 12, T09005 (2017),10.1088/1748-0221/12/09/T09005.

[22] A. Temnykh, D. Dale, E. Fontes, A. Lyn-daker, Y. Li, J. Ruff, P. Revesz, and A. Woll,“Chess upgrade with compact undulator mag-nets: Operating experience and first results,”AIP Conference Proceedings 1741, 020003 (2016),https://aip.scitation.org/doi/pdf/10.1063/1.4952782.

[23] P. Revesz, A. Lyndaker, A. Pauling, and A. Temnykh,“Novel video beam position monitor for canted undula-tors at CHESS,” CHESS E-News, 2015.

[24] Dmitry Teytelman, iGp-1281F Signal Processor: Techni-cal User Manual, Revision 1.7 , Tech. Rep. (Dimtel, Inc.,San Jose, CA, 2009).

[25] P. K. Lambert, C. J. Hustedt, K. S. Vecchio, E. L.Huskins, D. T. Casem, S. M. Gruner, M. W. Tate,H. T. Philipp, A. R. Woll, P. Purohit, J. T. Weiss,V. Kannan, K. T. Ramesh, P. Kenesei, J. S. Okasin-ski, J. Almer, M. Zhao, A. G. Ananiadis, and T. C.Hufnagel, “Time-resolved x-ray diffraction techniquesfor bulk polycrystalline materials under dynamic load-ing,” Review of Scientific Instruments 85, 093901 (2014),https://doi.org/10.1063/1.4893881.

[26] Jin Liu, Nicolas Dauphas, Mathieu Roskosz, Michael Y.Hu, Hong Yang, Wenli Bi, Jiyong Zhao, Esen E. Alp,Justin Y. Hu, and Jung-Fu Lin, “Iron isotopic fraction-ation between silicate mantle and metallic core at highpressure,” Nature Comm. 8, 14377 (2017).

[27] APS-U Preliminary Design Report, Tech. Rep. (ArgonneNational Laboratory).

[28] ESRF Upgrade Programme Phase II Technical DesignStudy , Tech. Rep. (European Synchrotron Radiation Fa-cility, 2015).

[29] F.J. Sacherer, “Bunch lengthening and microwave insta-bility,” IEEE Trans. Nucl. Sci. NS-24, 1393 (1977).

[30] Robert L. Gluckstern, Johannes van Zeijts, and BrunoZotter, “Coupling impedance of beam pipes of generalcross section,” Phys. Rev. E 47, 656–663 (1993).

[31] Advanced Computational Electromagnetics 3P CodeSuite, SLAC, Stanford, CA.

[32] L Lee, V Akcelik, L Ge, S Chen, G Schussman, A Candel,Z Li, L Xiao, A Kabel, R Uplenchwar, C Ng, and K Ko,

17

“Computational science research in support of petascaleelectromagnetic modeling,” Journal of Physics: Confer-ence Series 125, 012077 (2008).

[33] Laura Salo, Estimating the Logitudinal and TransverseIntensity-DependentBehavior of CESR, REU Report(CLASSE, Cornell University, Ithaca, NY, 2017).

[34] Joseph Raymond Calvey, Studies of Electron CloudGrowth and Mitigation at Cesr-TA, Ph.D. thesis, Cor-nell University, Ithaca, New York (2013).

[35] Keefer Rowan, Revised Predictions of Electron–Cloud-Based Instabilities for CHESS-U , REU Report(CLASSE, Cornell University, Ithaca, NY, 2018).